Beyond Wavelets and Deep Learning: A Model-Based Fusion Framework Using Discrete Band-Limited Shearlets

The limited dynamic range of standard digital imaging sensors often results in under- or over-exposed images, failing to capture the full radiance of a scene as perceived by the Human Visual System (HVS). While deep learning-based Multi-exposure Image Fusion (MEF) methods have recently dominated the field, they often suffer from dependencies on large datasets and lack interpretability. To address this, we propose a novel, model-based MEF framework leveraging the Discrete Band-Limited Shearlet Transform (DBLST). The shearlet transform provides a superior multi-scale and directional representation compared to traditional wavelets, making it exceptionally adept at capturing edges and textures across varying exposure levels—a crucial capability for high-quality fusion. Our method decomposes source images using DBLST and fuses the coefficients through specifically designed rules for low-pass and high-pass components. Extensive experiments on standard datasets demonstrate that the proposed algorithm not only significantly outperforms conventional wavelet-based methods but also achieves competitive performance against recent state-of-the-art approaches in terms of both subjective visual quality and objective metrics (including structural similarity, entropy, and average gradient). The results confirm DBLST as a powerful, efficient, and interpretable tool for high-dynamic-range image rendering, offering a compelling alternative to data-driven deep learning models.